Effects of Buffer Layer on Structural Properties of Nonpolar (11$\overline{2}$0)-Plane GaN Film

Abstract

Nonpolar (11$\overline{2}$0) a-plane GaN films were grown on semipolar (1$\overline{1}$02) r-plane sapphire substrates using various buffer layers within a low-pressure metal organic chemical vapor deposition system. The structural properties of nonpolar a-plane GaN films were intensively investigated by X-ray diffraction and Raman spectra measurements. A set of buffer layers were adopted from a GaN layer to a composite layer containing a multiple AlN layers and a gradually varied-Al-content AlGaN layer, the full width at half maximum of the X-ray rocking curves measured along the [0001] and [10$\overline{1}$0] directions of a-plane GaN were reduced by 35% and 37%, respectively. It was also found that the basal-plane stacking faults (BSFs) density can be effectively reduced by the heterogeneous interface introduced together with the composite buffer layer. An order of magnitude reduction in BSFs density, as low as 2.95 × 10⁴ cm⁻¹, and a pit-free surface morphology were achieved for the a-plane GaN film grown with the composite buffer layer, which is promising for the development of nonpolar GaN-based devices in the future.

1. Introduction

The light emitting and optoelectronic devices fabricated based on III-nitrides offer a wide spectral range varying from near-infrared (InN) to deep ultraviolet (AlN) [1,2,3]. Remarkable progress has been reported on the research of InGaN-based green and red light-emitting diodes (LEDs) [4,5,6,7], AlGaN-based deep ultraviolet (DUV) LEDs [8,9], GaN-based ultraviolet photodetectors [10], and AlGaN-based solar-blind photodetectors [11]. High-bandwidth light communication could be achieved based on the above visible light and ultraviolet LEDs [9,12]. Accompanied with these achievements, both the efficiency and the modulation bandwidth of III-nitride-based LEDs were limited by the built-in electric-field-induced quantum-confined Stark effect (QCSE) [13,14,15]. Although the QCSE can be suppressed/eliminated by growing semipolar/nonpolar GaN-based structures [16], the epitaxial growth of semipolar/nonpolar III-nitride structures was still a great challenge [17,18]. Therefore, the development of semipolar/nonpolar structures has lagged behind the development of polar structures.

Recently, high-bandwidth semipolar (20$\overline{2}$1)-plane GaN-based LED has been widely reported for visible-light communication [19,20] due to the successful fabrication of high-quality commercial semipolar (20$\overline{2}$1)-plane GaN templates. Meanwhile, a bandwidth of 540 MHz for (11$\overline{2}$2)-plane semipolar green LEDs had been demonstrated by growing an (11$\overline{2}$2)-plane GaN template whose crystalline quality was similar to or close to that of polar GaN [21]. The research on nonpolar AlGaN-based ultraviolet light-related structures [22,23] and InGaN-based visible light-related structures [24,25] has been investigated, but there is still much room for the investigation of devices based on nonpolar III-nitrides. Obviously, the successful fabrication of nonpolar GaN templates is the basis for the realization of nonpolar GaN-based devices. Although free-standing nonpolar GaN templates of comparable quality to polar GaN are currently available, they are cut at a specific angle from ultra-thick c-plane GaN, are typically only a few millimeters (~5 mm) width, and are difficult to commercialize.

Similar to the studies such as the SiN_x interlayer [26] and the NH₃ pulsed-flow growth technology [27] on the conventional c-plane polar III-nitride structures, various technologies such as buffer layers [28,29], in situ SiN_x interlayers [30], growth interruption [31], and substrate nitridation [32] have been extensively investigated to acquire nonpolar GaN films. The blocking effect of the SiN_x insertion layer on BSFs was verified by the TEM measurements reported by M.J. Kappers et al. [33]. It was found that a nonpolar GaN film with a pit-free and smooth surface could be achieved by using a composite buffer layer containing a low-temperature-grown AlN layer sandwiched in between two high-temperature-grown AlN layers [28]. R. Hao et al. reported that the defects in nonpolar a-plane GaN film could be reduced using either a SiN_x interlayer or a direct-growth technology (without a nucleation layer) [34]. In addition, a composite buffer layer consisting of a low-temperature-grown AlN layer, a low-temperature-grown pulsed AlN layer, and a low-temperature-grown AlGaN layer improved the crystal quality of nonpolar a-plane GaN film [35]. Recently, (11$\overline{2}$0)-plane nonpolar GaN film was successfully grown on a (0001)-plane sapphire substrate using pulsed laser deposition (PLD) technology [36]. It was reported that the crystalline quality of nonpolar GaN could be improved by introducing a patterned SiO₂ layer [37]. However, the introduction of secondary epitaxy processes or growth techniques that are not entirely based on the metal organic chemical vapor deposition (MOCVD) system is detrimental to commercialization. Therefore, epitaxial structures grown in a single process by using pure MOCVD technology are usually pursued.

In this study, several nonpolar (11$\overline{2}$0)-plane GaN films were grown on r-plane sapphire substrates by using a low-pressure MOCVD system. The buffer layer was modified to investigate the effect of the buffer layer on the structural properties of nonpolar (11$\overline{2}$0)-plane GaN films. It was found that a composite buffer layer containing multiple AlN layers and a gradually varied-Al-content AlGaN layer was effective in improving the crystal quality and suppressing the basal-plane stacking faults (BSFs) density. The FWHM of XRCs measured along the [0001] and [10$\overline{1}$0] directions of a-plane GaN was reduced by 35% and 37%, respectively, and a BSFs density as low as 2.95 × 10⁴ cm⁻¹ was achieved using the optimized composite buffer layer.

2. Experiment

In this study, a low-pressure MOCVD system was used to grow the nonpolar (11$\overline{2}$0) a-plane GaN films on the semipolar (1$\overline{1}$02) r-plane sapphire substrates. This MOCVD system is a domestically manufactured 2-inch monolithic system with a vertically coupled chamber. The chamber pressure was maintained at 40 Torr throughout the growth process. The precursors for Al, Ga, and N elements were trimethylaluminum (TMAl), trimethylgallium (TMGa), and ammonia (NH₃), respectively, while the carrier gas of the MOCVD system was hydrogen (H₂). Five nonpolar GaN samples, labeled as A–E, were grown based on various buffer layer structures to investigate the effects of the buffer layer on the structural properties of nonpolar GaN films. The epitaxial structures of the nonpolar GaN films are shown in Figure 1. Specifically, the buffer layer of sample A contained a low-temperature (650 °C) grown GaN nucleation (LT-GaN) layer and a high-temperature (1030 °C) grown GaN (HT-GaN) layer. The growth duration time, NH₃ flow rate, and TMGa flow rate of the LT-GaN layer were 7 min, 112 mmol/min, and 15.6 μmol/min, respectively, while these parameters were 15 min, 80 mmol/min, and 38.9 μmol/min, respectively, for the growth of the HT-GaN layer. For sample B, its buffer layer contained a low-temperature (720 °C) grown AlN layer (LT-AlN) and a high-temperature (1090 °C) grown AlN (HT-AlN) layer, respectively. Correspondingly, the growth duration time, NH₃ flow rate, and TMAl flow rate of the LT-AlN layer were 7 min, 62.5 mmol/min, and 3.5 μmol/min, respectively, while these parameters were 30 min, 44.6 mmol/min, and 23.2 μmol/min, respectively, for the growth of the HT-AlN layer. The buffer layer of sample C was a multiple AlN layer, which was identical to our previous report [30]. Specifically, it contained a 720 °C grown LT-AlN layer sandwiched in between two 1050 °C grown HT-AlN layers. The NH₃ and TMAl flow rates were 22.3 mmol/min, and 8.7 μmol/min for all the three AlN layers, respectively, and the growth duration time was 3, 4.5, and 7.5 min for the first HT-AlN layer, the LT-AlN layer, and the second HT-AlN layer, respectively. An AlGaN layer was grown on the multiple AlN layers of sample C to constitute the buffer layer of sample D. The growth temperature, duration time, NH₃ flow rate, TMAl flow rate, and TMGa flow rate were 1090 °C, 15 min, 112 mmol/min, 14.7 μmol/min, and 32.4 μmol/min, respectively. In other words, the gas phase Al content of this AlGaN layer was ~0.31. For structure E, the fixed Al content AlGaN layer was replaced by a varied-Al-content AlGaN layer whose gas phase Al content was gradually decreased from 0.7 to 0.07 from the beginning to the end, as a strain-relaxation layer. The growth temperature and NH₃ flow rate were fixed at 1090 °C and 84.8 mmol/min, respectively, while the flow rate of TMAl and TMGa was varied over the growth time. As shown in Figure 2, the flow rate of TMAl was started at 29.0 μmol/min, and gradually decreased linearly over time to 2.9 μmol/min, while the flow rate of TMGa was started at 12.3 μmol/min, and gradually increased linearly over time to 38.1 μmol/min. During this growth period, the flow rate of TMAl and TMGa was kept constant, so that the V/III ratio was maintained at ~2000.

After growing the buffer layer, an unintentionally doped nonpolar a-plane GaN film was grown on top of the buffer layer at 1030 °C. During this growth process, the flow rates of NH₃ and TMGa were 67.0 mmol/min and 68.1 μmol/min, respectively, resulting in a growth rate of 1.8 μm/h, as verified in our previous work [30]. In this study, a 1.2 μm thick a-plane GaN film was grown in 40 min. After growing all the nonpolar a-plane GaN samples, the crystalline quality was characterized by measuring the X-ray rocking curves (XRCs) of (11$\overline{2}$0)-plane GaN along the in-plane c- and m-directions. The BSFs density was calculated based on the modified Williamson–Hall analysis by measuring the XRCs of (h0$\overline{\mathrm{h}}$0)-planes GaN. Meanwhile, the in-plane strains along the c- and m-directions of nonpolar GaN were analyzed by measuring Raman spectra at RT under the x(-,-)$\overline{x}$ configuration. The surface morphology was observed with an optical microscope at 500X.

3. Results and Discussions

The X-ray diffraction (XRD) 2θ-ω scanning curves for samples A–E are shown in Figure 3a. The two main peaks located at 2θ = 52.56 and 57.76° were diffracted from the (2$\overline{2}$04)-plane of sapphire and (11$\overline{2}$0)-plane of GaN, as labeled in Figure 3a. These XRD 2θ-ω scanning curves indicated the successful growth of nonpolar a-plane GaN film. For samples B–E, a relatively weak peak located at 59.35° can be observed. These peaks were diffracted from the (11$\overline{2}$0)-plane AlN. On the other hand, a distinct “shoulder peak” can be distinguished between the XRD peaks of GaN and AlN, which was diffracted from the inserted AlGaN layers for samples D and E. The relatively broad diffraction peak of AlGaN in sample E was due to the fact that this layer was grown with a gradually decreasing gas phase Al content. Meanwhile, the XRCs were scanned along the in-plane [0001] c-direction and [10$\overline{1}$0] m-direction for each sample, and their corresponding full width at half maximum (FWHM), denoted as C_FWHM and M_FWHM, respectively, were determined by a single Gaussian function fitting and are illustrated in Figure 3b. These relatively large FWHM indicated that the crystalline quality of nonpolar GaN was inferior to that of conventional polar GaN, which was the main reason for the slow development of nonpolar GaN-based structures compared to their polar counterpart [17]. The fact that C_FWHM was much smaller than M_FWHM indicated the anisotropy in the crystalline quality of nonpolar GaN due to the anisotropy in lattice mismatch between the GaN epilayer and the sapphire substrate along the c- and m-directions [38], which is a typical feature of nonpolar GaN [18,30]. In fact, the in-plane c- and m-directions of the nonpolar a-plane III-nitrides epitaxially grown on an r-plane sapphire substrate are parallel to the [$\overline{1}$101] and [11$\overline{2}$0] directions of the sapphire substrate, respectively [38]. Here, the lattice mismatch is defined as follows:

$$\delta =\frac{a_{sub}-a_{epi}}{a_{epi}}\times 100\%,$$

(1)

where $a_{sub}$ and $a_{epi}$ are the lattice lengths along a specific direction of the substrate and epitaxial layers, respectively. Along the c-direction of nonpolar a-plane III-nitrides, the lattice lengths for GaN and AlN are 4.982 and 5.185 Å, respectively, i.e., their lattice constant c. The lattice length for [$\overline{1}$101] sapphire is 5.09 Å. In contrast, along the m-direction of nonpolar a-plane III-nitrides, the lattice lengths for GaN and AlN are 5.39 and 5.52 Å, respectively, which is $\sqrt{3}$ times their lattice constant a. The lattice length for [11$\overline{2}$0] sapphire is 4.71 Å [39]. Thus, the lattice mismatch along the two directions between the grown nonpolar III-nitrides and sapphire substrate can be calculated and are summarized in

Table 1. The lattice mismatch between the nonpolar a-plane III-nitrides and the semipolar r-plane sapphire substrate.

III-Nitrides	Along c-Direction of III-Nitrides	Along m-Direction of III-Nitrides
GaN	−1.83%	−14.73%
AlN	2.17%	−12.62%

.

The FWHM of the XRCs was found to vary with respect to the buffer layers. Among them, the crystalline quality of samples B and C contacted with the AlN buffer layers was slightly better than that of sample A containing only a GaN buffer layer. The improved quality was inferred due to the introduction of tensile stress along the c-direction induced by the AlN buffer layer, which partially compensated the compressive stress in the upper GaN layer [40]. Therefore, due to the extra stress relaxation effects of the AlGaN layer, the crystal quality of samples D and E was further improved, and the best crystal quality was achieved for sample E, which contained a gradually varied-Al-content AlGaN layer. Specifically, the C_FWHM and M_FWHM for sample E were reduced by 35% and 37%, respectively, compared to sample A, due to the introduction of the optimized buffer layer. The reduced C_FWHM and M_FWHM indicated the improved crystalline quality and the reduced structural defects, which will improve the performance of future devices.

The BSFs in nonpolar a-plane III-nitrides stacked along the c-direction and propagated from bottom to top along the growth direction, the a-direction, and acted as a leaky, non-radiative recombination center. Thus, the BSFs are usually considered to be the major defects in nonpolar III-nitrides. To investigate the effects of different buffer layers on the BSFs density, the modified Williamson–Hall analysis [41] was introduced to calculate the BSFs density in all the five a-plane nonpolar GaN samples. First, the XRCs of the (10$\overline{1}$0)- and (20$\overline{2}$0)-planes of each nonpolar GaN sample were measured along the c-direction. Then, the BSFs density was derived based on the flow equation:

$$\mathsf{\Delta }{\omega }_{\mathrm{measured}}=\mathsf{\Delta }{\omega }_{\mathrm{mosaic}}+\frac{\lambda }{2L\mathit{sin}({\theta }_{hkl})}.$$

(2)

Here, 1/L is the BSFs density, which can be calculated by a linear fitting based on the FWHM values of the XRCs diffracted from the (10$\overline{1}$0)- and (20$\overline{2}$0)-planes of GaN. The XRD ω-2θ scanning curves of the (10$\overline{1}$0)- and (20$\overline{2}$0)-planes for sample A are shown in Figure 4a as representative. Figure 4b demonstrates the modified Williamson–Hall plots for the five samples based on the measured XRCs. The intercept of the corresponding fitted line in Figure 4b is the calculated BSFs density for each sample, and their values are plotted in Figure 4c.

It can be seen from Figure 4c that the BSFs density in sample A was the highest among the five samples, while the BSFs density in samples B and C was slightly decreased after the use of the AlN buffer layer. After the introduction of the AlGaN layer, the BSFs density in sample D was significantly reduced by about 40% compared to samples B and C, and by 54% compared to sample A. When the gradually varied-Al-content AlGaN layer was introduced for sample E, the BSFs density was further reduced by 71% compared to sample D, as low as 2.95 × 10⁴ cm⁻¹, and an order of magnitude reduction was achieved compared to sample A. This density was reduced by 64% compared to our previous work (8.13 × 10⁴ cm⁻¹) on the growth of nonpolar a-plane GaN using an in situ SiN_x interlayer [42], and reduced by 56% compared to the result (6.7 × 10⁴ cm⁻¹) reported by J. Liang et al. on the growth of a-plane GaN on lattice-matched LaAlO₃ substrates [43]. In fact, the evolution trend of the BSFs density of the five samples was basically consistent with that of the FWHM of the XRCs, as shown in Figure 3b. In general, the BSFs, which were formed at the initial stage of growth and propagated along the growth direction, can be partially blocked by heterogeneous interfaces. The decrease in the BSFs density in samples B and C compared to sample A could be attributed to the blocking of BSFs by the AlN buffer/upper GaN interface. For sample D, two additional interfaces, the AlN/AlGaN and AlGaN/GaN interfaces, were added based on sample C, both of which can have the effect of blocking the propagation of BSFs. The varied-Al-content AlGaN layer in sample E corresponded to many heterogeneous interfaces, exerting a continuous blocking effect on the propagation of BSFs. Therefore, a significant (about an order of magnitude) decrease in the BSFs density was achieved for sample E compared to sample A, proving that the gradually varied-Al-content AlGaN layer played a major role in suppressing the BSFs density.

The Raman spectra measured at room temperature under a configuration of x(-,-)$\overline{x}$ for samples A–E are plotted in Figure 5a. The three Raman peaks marked with * were sapphire peaks [44], and the right peak marked with # was the overlapping peak of sapphire and GaN E₁(LO) [44,45]. The three Raman peaks located between 500 and 600 cm⁻¹ from left to right were the GaN A₁ (TO), E₁ (TO), and E₂ (high) mode-related Raman peaks, respectively, as marked in Figure 5a. The in-plane strains of nonpolar GaN films can be calculated based on the Raman peak positions of the E₁ (TO) and E₂ (high) modes. In fact, the mathematical relationship between the strains along different GaN in-plane directions and the Raman shift of the corresponding mode is represented by the following equation:

$$∆{\omega }_{\lambda }=a_{\lambda }\left(1-\frac{c_{12}}{c_{11}}\right){\epsilon }_{yy}+\left(b_{\lambda }-a_{\lambda }\frac{c_{13}}{c_{11}}\right){\epsilon }_{zz}+c_{\lambda }\left|{\epsilon }_{xx}-{\epsilon }_{yy}\right|.$$

(3)

Here, $∆{\omega }_{\lambda }$ represents the Raman shift of a particular mode, c₁₁, c₁₂, and c₁₃ are the elastic stiffness constants for GaN [46], $a_{\lambda }$, $b_{\lambda }$, and $c_{\lambda }$ represent the corresponding phonon deformation potentials for the particular mode [38,47], and ${\epsilon }_{xx}$, ${\epsilon }_{yy}$, and ${\epsilon }_{zz}$ are the strains along the GaN $\left[11\overline{2}0\right]$, $\left[1\overline{1}00\right]$, and [0001] directions, respectively [32,48]. The peak positions for the E₁(TO) and E₂(high) modes of strain-free GaN were assigned to be 558.8 and 567.6 cm⁻¹, respectively. The measured peak positions of the E₁(TO) and E₂(high) modes of samples A–E are summarized in

Table 2. The measured peak positions of E₁(TO) and E₂(high) modes of nonpolar GaN samples A–E.

Raman Mode	Sample A	Sample B	Sample C	Sample D	Sample E
E1(TO), cm−1	556.11	556.40	554.92	557.04	558.45
E2(high), cm−1	565.13	565.79	564.13	566.29	567.55

. Then, two equations for the strains along different directions can be obtained by substituting the Raman shifts of the E₁ (TO), and E₂ (high) modes into the above equation. Meanwhile, the strain along the growth direction is assumed to be zero based on the linear elasticity theory, while the strains, ${\epsilon }_{xx}$, ${\epsilon }_{yy}$, and ${\epsilon }_{zz}$ are subject to the following expressions [38]:

$${\epsilon }_{yy}=-\frac{c_{11}}{c_{12}}{\epsilon }_{xx}-\frac{c_{13}}{c_{12}}{\epsilon }_{zz}.$$

(4)

Then, a ternary system of equations was constructed and the in-plane strains, ${\epsilon }_{zz}$ and ${\epsilon }_{yy}$, were obtained by solving the above equations. The solved in-plane strains of the five nonpolar GaN are demonstrated in Figure 5b.

The in-plane strains in sample A grown on GaN buffer were significantly reduced compared to those in samples B and C grown on AlN buffer. It can be concluded that the in-plane strains in sample A were partially released by the formation of BSFs at the initial stage of growth. For samples B and C, the strain along the c-direction was partially compensated by the tensile strain introduced along with the AlN buffer layer. In this case, the residual strain in the GaN layer was reduced, and therefore the proportion of strains released by formatting the BSFs was synchronously reduced. As a result, the in-plane strains in samples B and C were larger, and the BSFs density was smaller than those in sample A.

When an additional layer of AlGaN was introduced for samples D and E, the in-plane strains were slightly reduced due to the lattice mismatch between the top GaN and bottom AlN being transited by the inserted AlGaN layer. In other words, due to the lattice lengths along a certain direction of the AlGaN being between AlN and GaN, the lattice mismatch of the GaN film for samples D and E was between the upper GaN layer and the lower AlGaN layer, while the lattice mismatch of the GaN film for samples B and C was between the upper GaN layer and the lower AlN layer. This means that when the AlGaN was introduced, the lattice mismatch between the GaN film and the lower layer was actually reduced. On the other hand, the propagation of the BSFs could be partially blocked by the additional interfaces introduced along with the AlGaN layer. Thus, both the FWHM of the XRCs and the BSFs density were improved with the introduction of the AlGaN layer. Obviously, the AlGaN layer whose Al content varied gradually from 0.7 to 0.07 in sample E produced a better transition effect on the lattice mismatch, so that sample E achieved a smaller in-plane strains compared to sample D.

As shown in Figure 6, the directional stripes distributed on the surface parallel to the in-plane c-direction are due to the fact that the in-plane growth rate along the c-direction was higher than that along the m-direction [49], and are the typical surface morphology of (11$\overline{2}$0)-plane GaN-based materials [25]. It can be seen from Figure 6 that relatively more triangular pits were distributed on the surface of samples B and C, while the distribution on the surface of sample C was less, and samples A and E achieved a pit-free surface morphology. The formation of these triangular pits is usually related to the strains in nonpolar GaN. The pit-free surface of samples A and E was inferred because the strains were partially released by the formation of BSFs for sample A, and were compensated by the introduced composite buffer layer for sample E. Therefore, the composite buffer layer introduced in this study can improve the crystal quality, reduce the BSFs density, and form a pit-free surface in the growth of nonpolar a-plane GaN films.

4. Conclusions

The effects of buffer layers on the structural properties of nonpolar (11$\overline{2}$0) a-plane GaN films have been intensively investigated. The in-plane strains can be suppressed by optimizing the buffer layers, resulting in improved crystalline quality. An AlGaN layer was introduced to compensate for the in-plane strains in the a-plane GaN due to the usage of AlN buffer layer. It was calculated that the C_FWHM and M_FWHM were reduced by 35% and 37%, respectively, compared to a reference sample using GaN as the buffer layer. Meanwhile, the BSFs density can be effectively reduced by the heterogeneous interface introduced together with the buffer layers. A BSFs density as low as 2.95 × 10⁴ cm⁻¹, an order of magnitude lower than that of the reference sample, and a pit-free surface morphology were achieved for the sample grown with a composite buffer layer containing multiple AlN and gradually varied-Al-content AlGaN layers, providing promising insights for the development of nonpolar devices in the future.